Bank lending and commercial property cycles: some cross

Download Report

Transcript Bank lending and commercial property cycles: some cross

BANK LENDING, BANK
PERFORMANCE AND
COMMERCIAL PROPERTY
PRICES
Course on Financial Instability at the Estonian Central Bank,
9-11 December 2009 – Lecture 9
E Philip Davis
NIESR and Brunel University
West London
[email protected]
www.ephilipdavis.com
groups.yahoo.com/group/financial_stability
PAPER 1:
BANK LENDING AND
COMMERCIAL PROPERTY
PRICES:
some cross-country evidence
E Philip Davis and Haibin Zhu
Revise and resubmit in Journal of
International Money and Finance
Introduction
• Growing interest in commercial property cycles and
link to financial stability
• Likely to be more volatile than residential given no
intrinsic reservation value
• Key role of banks in financing commercial property,
while CP is also widely used as collateral for non-CP
lending
• Little empirical evidence on link from commercial
property cycle to credit cycle, notably at international
level
Literature review
• Explanations of real estate cycles
– Value determined by discounted future rents and
investment by a valuation ratio
– Distinctive features of asset market including
heterogeneity, lack of central trading, high
transactions costs, supply constraints…
– …and use as collateral for bank loans…
– …while external financing needed for construction
and occupancy – generally bank debt
– So optimism raising demand can drive up prices
while supply response slow - when supply comes
on stream may be excessive relative to demand,
driving prices down
– Traditionally such a pattern is seen as requiring not
just sticky supplies and rents but also irrationality
– basing expected profitability of construction on
current prices
– Examples are rules of thumb, myopic expectations,
disaster myopia
– Some urge cycles impossible with rational
expectations, but following are possible “rational”
causes:
• No short selling possible to stabilise market
• Option value of investment in “anticipated
uncertainty”
• Long leases and use of credit
• Collateral effects on borrowing capacity,
including the “financial accelerator”
• Risk shifting behaviour by banks
– Empirical work in “real estate” literature illustrates
interaction of investment, rents and prices, as well
as scope for bubbles
• Property prices and bank lending
– Background: commercial property price booms
and busts preceding banking crises. Three
dimensions of interaction:
(i) Reasons property prices affect credit
• Investment channel
• Wealth effect on borrowers boosting credit
demand
• Banks ownership of property boosting capital
base increases banks’ lending capacity
• Financial accelerator effect making lending
procyclical, especially if default risk
underestimated in booms
(ii) Reasons lending could affect property prices
• Liquidity effect
• Credit raising real estate demand; short term
positive effect
• Credit raising real estate supply; long term
negative effect
• Supply of credit boosted when banks compete,
e.g. after financial liberalisation
• Directed to real estate if high quality borrowers
shift to securities market or internal finance
• Aggravated by moral hazard
(iii) Common economic factors for lending and real
estate prices
• Credit affected by shocks to variables such as
GDP and interest rates…
• …which also provoke demand and supply
imbalances in real estate
(iv) Will changing nature of finance affect the creditproperty price interrelation?
• Note in particular that in financially-liberalised
regime, effect of credit on prices is less likely
(lending accomodates to demand rather than
being rationed, while prices adjust in forward
looking manner)
• Extant empirical work
– Country-specific studies of interaction with
banking system…
– …international studies mainly use residential or
mixed prices, including prediction of financial
instability
– But no major academic research project has yet
looked at threats to financial stability from the
commercial property sector on a systematic,
empirical, cross-country basis. This is an important
motivation for our own work.
A model of real estate cycles (based on
Carey and Wheaton)
Economic environment
– N investors
– Heterogeneous valuation of properties, with a
distribution of F(P)
– Banks’ lending attitude varies over time wt
– Bank lending function for investors: L(Y, i, P, wt)
– Supply K is fixed in short run but adjusts slowly in
response to prices exceeding replacement cost, with
separate lending function B(Y,I,P,wt)
– Investment depends on current property prices, for
reasons set out above – irrationality, bank capital
effects and credit market imperfections
Model
• Market demand function (1), supply
adjustment (2), new investment (3) and market
clearing (4)
N [1  F ( Pt )] L(Yt , it , Pt , wt )
Dt 
,
Pt
LY  0, Li  0, LP  0
K t  (1   ) K t 1  I t 1
I t 1    Bt 1 (Yt 1 , it 1 , Pt 1 , wt 1 ),
Dt  K t
(1)
(2)
BY  0, Bi  0, B P  0
(3)
(4)
• Relationship between property prices and bank
lending (Lt+Bt)
– Higher current property prices increase bank
lending
– Higher Lt (e.g. due to financial liberalisation w)
increases current property prices
– Higher Bt reduces future property prices
– Both affected by macroeconomic factors (Y, i)
• Simplification – 2 equations, 2 unknowns (K, P)
*
*
N
[
1

F
(
P
)]
L
(
Y
,
i
,
P
, w)
*
K 
P*
  K *    B * (Y , i , P * , w )
(5)
(6)
• Hypothesis I: (collateral/financial accelerator effect) An
increase in commercial property prices has a positive impact on
bank credit.
• Hypothesis II: (liquidity effect) Bank credit can have offsetting
impacts on commercial property prices. New credit to the
demand (investor) side may increase property prices in the short
run, while new lending to the supply (constructor) side may tend
to reduce property prices in the long run.
• Hypothesis III: (macro effect) Commercial property prices
adjust to changes in macroeconomic conditions. Their dynamic
adjustment depends on the characteristics of the property market
in each country. In particular, if the supply is more elastic than
the demand, the market reacts to a macro shock in the form of an
oscillation around the new steady state; otherwise property prices
“overshoot” and then gradually converge to the new steady state.
Empirical analysis
• Data
– 17 countries: Australia, Belgium, Canada,
Denmark, Finland, France, Germany, Ireland, Italy,
Japan, Netherlands, Norway, Spain, Sweden,
Switzerland, the UK and the US
– Main focus interrelation of real commercial
property prices, GDP, investment, real credit and
real short rates
– Most countries’ “true” data is annual – mainly used
in our work
– Stationarity as preliminary – all have unit root
except real short rate
• Determination of commercial property prices
Error Correction estimation
– Panel estimation, GLS, cross section weights, White standard
errors. ECM tends to be highly significant
– For all countries:
• Strong short run effect of GDP and credit growth – implies
high cyclical volatility – consistent with model
• Long run positive link to GDP and negative to credit –
plausible in terms of model
• Positive real short rate – financial liberalisation?
– Subgroups
• G-7, SOEs, bank and market oriented, crisis countries
broadly similar to full panel
• Main contrast is with crisis countries over 1985-95 – long run
positive credit and negative investment effect, very high short
run elasticities
Results of panel estimation
Pooled
G-7
Small
open
Economies
Bank
dominated
Market
oriented
Crisis
countries
Fixed
effect
DLCREDR 0.75
(6.4)
DLGDP
1.78
(6.3)
DLI
Fixed
effect
0.92
(5.5)
1.13
(2.8)
Fixed
effect
0.67
(3.0)
1.8
(4.3)
-0.04
(2.2)
-0.08
(2.2)
Fixed
effect
0.84
(5.2)
1.47
(2.9)
0.29
(1.8)
-0.16
(6.1)
Fixed
effect
1.22
(9.9)
LCPPR(-1)
Fixed
effect
0.71
(4.6)
1.14
(2.3)
0.28
(5.0)
-0.13
(5.0)
-0.04
(1.9)
-0.073
(2.1)
-0.093
(4.0)
-0.14
(2.3)
0.21
(2.3)
Constant
LCREDR(1)
LGDP(-1)
-0.09
(5.3)
-0.09
(2.4)
0.17
(2.3)
LI(-1)
RSR
RSR(-1)
R-bar-sq
SE
DW
OBS
0.067
(1.8)
0.15
(2.5)
0.005
(2.4)
0.35
0.11
1.37
439
0.39
0.098
1.23
185
0.096
(1.7)
0.004
(1.6)
0.32
0.12
1.42
239
0.005
(1.9)
0.34
0.11
1.54
285
0.51
0.09
1.00
126
0.38
0.11
1.35
201
Crisis
countries
19851995
Fixed
effect
2.4
(5.4)
3.5
(3.8)
-0.88
(3.7)
-0.18
(4.3)
0.4
(2.9)
All
countri
es
19851995
Fixed
effect
1.2
(7.7)
2.18
(4.7)
-0.66
(3.7)
-0.31
(5.6)
0.64
0.11
1.66
88
0.008
(1.8)
0.65
0.11
1.52
194
-0.13
(4.0)
Interaction between bank lending and
commercial property prices
• Above evidence gives no view on causality links
between credit, commercial property prices and
macroeconomic fundamentals
• Granger causality suggests that commercial property
prices most commonly precede credit (9 countries)
(possibly via effects on collateral and capital), but
some reverse causality and interactions (7 countries)
• Granger causality needs supplementing as only
bivariate
• Test for dynamic interaction
• Method: VECM if there exists cointegration
(Johansen); VAR otherwise (CA, FI, IT, DK, NO,
CH)
• Endogeneity issue
• Need for choice of recursive ordering in order to
undertake Choleski decomposition
• Preferred ordering GDP, commercial property prices,
credit, investment, real short rates
• GDP first and interest rate last reflects transmission
mechanism lags
• Investment after credit and prices due to supply lags
• Prices before credit reflects role of collateral and
price stickiness
• Variance decomposition shows autonomy of
commercial property prices (47% in 5 years)
• Link to credit only significant in BE, IT, SE and CH suggests Granger Causality suffered omitted variables
bias
• Wider range of countries show link to GDP – main
external influence on commercial property prices
• Credit less autonomous, main influences on variance
are GDP (33%) and commercial property prices
(20%)
• Overall, confirms influence of external shocks (GDP)
on the nexus and of prices on credit
• Variants largely confirm these results
VECM variance decomposition
Real commercial property prices
GDP
CPP
CRED
I
Real private sector credit
RSR
GDP
CPP
CRED
I
RSR
Australia
Belgium
Canada
Denmark
Finland
France
Germany
Ireland
Italy
Japan
Netherlands
Norway
Spain
Sweden
Switzerland
UK
US
40
41
Na
56
Na
38
11
14
Na
10
11
29
9
32
7
17
43
40
28
Na
34
Na
52
83
44
Na
76
47
66
16
44
40
67
18
12
28
Na
3
Na
3
2
9
Na
1
13
3
18
22
46
1
1
1
1
Na
5
Na
6
3
6
Na
2
24
1
53
0
5
11
30
7
2
Na
1
Na
0
1
26
Na
11
3
2
5
0
2
4
8
75
1
Na
66
Na
55
10
37
Na
31
14
46
28
20
1
31
42
9
2
Na
2
Na
23
45
23
Na
29
49
32
3
19
3
35
11
11
85
Na
20
Na
6
11
3
Na
4
27
21
68
58
94
31
28
0
11
Na
7
Na
13
8
14
Na
10
1
1
4
2
1
4
13
5
1
Na
6
Na
3
27
28
Na
26
9
0
0
1
1
0
7
Mean level
26
47
12
11
5
33
20
33
6
8
Memo:
without RSR
18
58
11
14
34
19
43
4
Memo:
lags
1
1
Na
1
Na
1
1
1
Na
1
1
1
1
1
1
1
1
• Impulse response function
– Response of CPP to credit: positive short-term
effect but negative long-term impact in most
countries – consistent with theory.
– Response of CPP to GDP: differ by characteristics
of national markets. Two types of responses:
– Overshooting in 9 countries (Australia is a typical
case)
– Oscillation in 5 countries
Impulse response of prices to credit
GERMANY
DENMARK
Response of DELCPPR to Cholesky
One S.D. DELCREDR Innovation
Response of DKLCPPR to Cholesky
One S.D. DKLCREDR Innovation
.01
.02
.01
.00
.00
-.01
-.01
-.02
-.02
-.03
-.03
-.04
-.05
-.04
-.06
-.05
-.07
1
2
3
4
5
6
7
8
9
10
1
2
3
UNITED KINGDOM
4
5
6
7
8
9
10
UNITED STATES
Response of UKLCPPR to Cholesky
One S.D. UKLCREDR Innovation
Response of USLCPPR to Cholesky
One S.D. USLCREDR Innovation
.020
.015
.015
.010
.010
.005
.005
.000
.000
-.005
-.005
-.010
-.010
1
2
3
4
5
6
7
8
9
10
-.015
1
2
3
4
5
6
7
8
9
10
Impulse response of prices to GDP
AUSTRALIA
GERMANY
Response of DELCPPR to Cholesky
One S.D. DELGDP Innovation
Response of AULCPPR to Cholesky
One S.D. AULGDP Innovation
.14
.070
.12
.065
.10
.060
.08
.055
.06
.050
.04
.045
.040
.02
1
2
3
4
5
6
7
8
9
1
10
2
ITALY
3
4
5
6
7
8
9
10
9
10
UNITED KINGDOM
Response of ITLCPPR to Cholesky
One S.D. ITLGDP Innovation
Response of UKLCPPR to Cholesky
One S.D. UKLGDP Innovation
.09
.16
.08
.14
.07
.12
.06
.05
.10
.04
.08
.03
.06
.02
.04
.01
.00
.02
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
Conclusions
• Presented a theoretical model which shows cycles
emerge under plausible assumptions and generating
predictions for effects of GDP, interest rates and
credit
• Commercial property prices show degree of
autonomy, link to GDP but influence on credit
• Predominant direction of causality is from CPP to
credit rather than vice versa – collateral/financial
accelerator and not liquidity effect; latter effect
possibly dampened as financial liberalisation
• Important effect of GDP on both CPP and credit.
• Policy aspects include:
– Collateral-based amplification: bank credit policy
• Maximum LTV
• Portfolio limits on loan concentration
• Valuation method: long run view of valuation vs.
current market value
– Financial crises caused by real-estate bubbles
– Further research needed
• effects of property prices on bank profitability at
micro level – paper 2
• Can commercial property prices predict banking
crises – research to be pursued
PAPER 2:
COMMERCIAL PROPERTY
PRICES AND BANK
PERFORMANCE
E Philip Davis and Haibin Zhu
Published in Quarterly Review of
Economics and Finance
Introduction
• Role of asset prices in bank lending and bank
performance
• Particular role of commercial property prices, as
witness major differences in bank behaviour and
performance during the up- and downswings in
commercial property prices
• Extensive macro work on commercial property
prices and lending (paper 1), but less micro
estimation on lending and performance
• Is there a direct impact on the lending decisions, risk
and profitability of individual banks?
Table 1
Bank lending and bank performance at different stages of commercial property cycles
(1979-2001)
Country
Growth rate of
bank loans (%)
Growth rate of
risk-weighted
assets (%)
Provisions on
loans as a
percentage of
net income (%)
Memo: number
of years
Up
swing1
Down
swing
Up
swing
Down
swing
Up
swing
Down
swing
Up
swing
Down
swing
Up
swing
Down
swing
Belgium
8.69
4.75
7.86
3.42
0.38
0.34
17.13
21.36
14
9
Canada
6.51
8.16
--
--
1.00
1.01
32.33
34.89
9
7
Finland
11.02
-1.73
--
--
0.21
0.32
37.02
27.95
18
5
France
7.42
2.67
--
--
0.44
0.27
30.63
58.25
14
9
Germany
7.33
8.58
--
--
0.54
0.59
39.79
41.44
14
9
Italy
13.02
7.77
9.19
3.29
1.04
0.70
25.73
37.97
8
10
Japan
12.34
-0.18
--
-8.87
0.48
-0.08
6.98
57.02
12
11
Netherlands
13.25
10.20
13.62
5.89
0.69
0.58
18.84
24.69
15
8
Norway
15.00
10.03
9.59
-0.13
0.94
0.02
23.32
145.92
14
9
Sweden
11.39
8.41
5.26
8.26
0.73
0.74
56.10
40.87
16
7
Switzerland
8.58
4.70
3.47
1.17
0.68
0.57
--
--
11
12
UK
10.48
10.45
9.74
14.68
1.02
0.85
--
--
11
12
US
9.64
5.07
9.59
3.62
1.39
1.17
22.59
39.52
9
14
10.36
6.07
8.54
3.48
0.73
0.55
28.22
48.17
Average
1
Return on
assets (%)
“Up (down) swing” refers to the years when real commercial property prices in that country increase (decrease).
Sources: OECD; BIS; authors’ calculations.
• We analyse a sample of 904 banks worldwide over the
period 1989-2002.
• Seek to assess the effect of changes in commercial
property prices on bank behaviour and performance in
a range of industrialised economies, focusing on
determination of lending, margins, ROA, bad debts
and provisioning
• Consistent with macro-level studies, commercial
property prices have a marked impact on the behaviour
and performance of individual banks, over and above
conventional determinants
• Results have implications for risk managers, regulators
and monetary policy makers.
Table 2
Distribution of sample banks
By country
Number of banks
By specialisation
Number of banks
Belgium
19
Bank holding company
428
Canada
21
Commercial bank
269
Finland
4
Cooperative bank
67
France
58
36
Germany
40
Investment bank /
securities house
Hong Kong
13
Median and long term
credit bank
12
Italy
38
26
Japan
143
Non-banking credit
institution
Real estate / Mortgage
bank
37
Savings bank
29
Total
904
Netherlands
8
Norway
14
Singapore
5
Sweden
5
Switzerland
28
United Kingdom
54
United States
454
Total
904
• Micro work – empirical analysis
– Provisioning (Laeven and Majnoni)
– Bank profitability and margins (Demirgüç-Kunt
and Huizinga)
– Bad loan ratios (Salas and Saurina)
– Lending (Bikker and Hu)
– Rare studies looking at CPP and bank
performance
•
•
•
•
Austria (Arpa et al)
Japan (Gan)
Hong Kong (Gerlach et al)
US (Hancock and Wilcox)
• Our advance on earlier literature
Empirical
work
– First international
study on
how commercial
property price movements affect individual
banks’ lending strategies and performance after
we control for the effects of conventional
explanatory variables (macro factors, bankspecific variables and country-specific factors)
– Micro-level data allow us to examine whether
the determination of bank performance and the
role of commercial property prices vary across
different groups of banks and across countries.
– Examine whether commercial real estate booms
and busts tend to have asymmetric impacts on
bank performance.
• Use of panel GLS or GMM (robustness
check)
• Control variables
– Macro: growth rate of real GDP, inflation and
short-term interest rates
– Bank: loan-to-asset ratios, real loan growth rate,
capital strength, net interest margin, bank size
dummies
– Country dummies
– Growth of real commercial property prices
Issues of endogeneity
• Basic GLS equations ignore dynamic interaction of
variables
– No lagged dependent variable
– Bank specific variables lagged
– Nationwide CPP likely to be exogenous to lending
behaviour of individual bank
– Previous results showed CPP largely autonomous of credit
even at macro level
– Major loss of observations
• Robustness checks
– Using lagged CPP
– Using difference and levels GMM estimation
Table 3
Summary statistics of regression variables
Variables
No. Obs
Mean (%)
Std. Dev. (%)
Min (%)
Max (%)
Asset growth
rate
5244
8.13
10.90
-49.17
49.72
Loan growth
rate
5132
8.54
12.03
-49.98
49.98
Loan to asset
ratio
6025
61.07
15.22
11.27
89.86
Net Interest
Margin (NIM)
5980
3.39
2.19
-5.88
36.72
Non-Performing
Loan ratio
(NPL)
4353
2.44
3.91
0.00
45.79
Return on
Assets (ROA)
6056
0.85
0.90
-7.65
8.79
Provisions /
Total Assets
5844
0.40
0.65
-2.16
16.36
GDP growth
rate
12656
2.44
2.11
-7.85
15.57
Inflation
12656
2.57
1.66
-4.04
10.97
Interest rate
12656
5.22
2.83
0.09
14.76
Growth rate of
real commercial
property prices
12651
-3.94
10.85
-49.19
35.49
Table 4
Characteristics of banks grouped by sizes1
Large banks
Mid-sized banks
Small banks
Variables
Mean
Loan growth rate
Std dev
Mean
Std dev
Mean
Std dev
5.91
10.36
5.45
11.90
9.12
12.11
54.79
14.49
62.33
14.93
61.52
15.19
NIM
1.82
0.86
2.13
1.45
3.67
2.23
NPL
4.58
4.06
4.34
6.23
2.15
3.58
ROA
0.37
0.58
0.44
0.81
0.94
0.91
Loan to asset ratio
1
There are 62 large banks, 76 mid-sized banks and 766 small banks.
Pooled regression with random effects
Dependent variables
Loan growth
rate
NIM
8.8***
(6.1)
1.94***
(7.8)
GDP growth
0.44***
(5.2)
0.05***
(10.6)
Inflation
-0.18
(0.9)
Interest rate
Constant
NPL
ROA
Provisions/
Total Assets
1.4**
(2.4)
0.42***
(3.2)
-0.21**
(2.4)
-0.046**
(2.2)
0.026***
(3.7)
-0.013***
(2.8)
Macro indicators
0.007
(0.6)
-0.58***
(10.2)
0.14***
(8.6)
-0.048***
(4.4)
0.42***
(4.0)
0.07***
(11.0)
0.12***
(4.4)
-0.053***
(5.8)
0.007
(1.2)
-0.083***
(5.6)
0.01***
(6.3)
-0.0023
(0.4)
-0.0057***
(4.1)
0.0037***
(4.1)
-0.0028***
(3.3)
-0.022***
(6.6)
0.0053***
(4.6)
-0.0043***
(5.6)
Bank indicators
Loan/Asset (-1)
Loan growth rate (-1)
NIM (-1)
0.47***
(3.6)
Capital ratio (-1)
0.084
(1.3)
0.14**
(2.5)
0.053***
(8.5)
-0.114***
(5.2)
0.27***
(23.4)
0.052***
(8.9)
EBTDA/Total assets (-1)
0.007*
(8.7)
0.0066*
(1.6)
0.06***
(5.6)
SMALL
Insig
0.74***
(3.5)
LARGE
Insig
Insig
1.0**
(2.5)
Insig
-0.25***
(3.2)
Insig
-0.11**
(2.3)
Insig
Commercial property
sector
D(CPP)
No. Obs.
0.16***
(9.4)
5052
-0.0095***
(8.8)
4195
-0.02***
(4.0)
3069
0.0095***
(6.1)
4182
-0.0049***
(4.8)
4060
Pooled regression with random effects and leveraged size effects
Dependent variables
Loan growth
rate
NIM
NPL
ROA
SMALL
Insig
0.42*
(1.9)
1.1**
(2.3)
-0.23**
(2.0)
-0.29**
(2.3)
LARGE
Insig
Insig
Insig
Insig
Insig
GDP*SMALL
Insig
Insig
-0.24***
(4.1)
Insig
0.022*
(1.6)
GDP*LARGE
Insig
Insig
-0.16*
(1.9)
Insig
Insig
IR*SMALL
Insig
0.08**
(3.8)
Insig
Insig
0.043**
(2.3)
IR*LARGE
Insig
Insig
Insig
Insig
Insig
INF*SMALL
-0.94*
(1.6)
Insig
Insig
Insig
Insig
INF*LARGE
Insig
Insig
Insig
Insig
Insig
D(CPP)
0.26***
(5.1)
-0.01***
(3.1)
D(CPP)*SMALL
-0.11**
(2.2)
Insig
D(CPP)*LARGE
Insig
No. Obs.
5052
0.0082*
(1.8)
4195
Provisions/
Total Assets
-0.053***
(3.4)
0.019***
(4.0)
-0.0168***
(5.6)
0.04**
(2.3)
-0.011**
(2.2)
0.014***
(4.4)
Insig
Insig
Insig
3069
4182
4060
Variants and robustness checks (1)
Dependent variables
Loan growth
rate
NIM
NPL
ROA
Provisions/
Total Assets
Real residential prices
DRRP
0.22***
(6.1)
-0.0285***
(14.3)
-0.094***
(7.3)
0.019***
(5.6)
-0.014***
(6.5)
0.065***
(7.0)
0.00184***
(3.7)
0.01***
(3.8)
0.0028***
(3.5)
-0.0002
(0.3)
DRCP
0.149***
(8.0)
-0.004***
(3.8)
-0.002
(0.4)
0.007***
(3.9)
-0.0028**
(2.5)
DRRP
0.05
(1.2)
-0.0245***
(11.4)
-0.09***
(6.1)
0.012***
(3.1)
-0.011***
(4.5)
0.055***
(5.0)
-0.0012***
(15.0)
-0.032***
(7.7)
0.006***
(4.6)
-0.0047***
(5.5)
-0.0092***
(9.6)
-0.02***
(3.9)
0.01***
(6.3)
-0.0055***
(5.3)
Real equity prices
DREP
Real residential and
commercial prices
Lagged commercial
prices
DRCPP(-1)
Nominal commercial
prices
DCPP
0.17***
(10.0)
Pooled regression with difference specification and lagged dependent variables (GMMdifference estimation)
Dependent
variables
Loan growth
rate
NIM
NPL
ROA
Provisions/
Total Assets
GMM difference
D.Lagged
variable
D.D(CPP)
0.053*
(1.8)
0.125***
(4.7)
0.77***
(6.7)
0.0
(0.6)
0.69***
(6.3)
-0.017***
(2.8)
0.35***
(4.5)
0.0058***
(3.3)
-0.014
(0.9)
-0.0037***
(2.7)
Observations
3305
3301
2250
3302
3225
Joint Wald
113 [0.0]***
119 [0.0]***
87 [0.0]***
61 [0.0]***
54 [0.0]***
Sargan
301 [0.47]
393 [0.41]
212 [1.0]
362 [0.22]
351 [0.12]
AR(1)
-7.7[0.0]***
-3.7 [0.0]***
-2.3 [0.02]**
-3.3 [0.001]***
-1.8 [0.08]*
AR(2)
-0.37 [0.72]
0.02 [0.98]
0.13 [0.9]
-1.8 [0.08]*
-1.7 [0.07]*
Pooled regression with lagged dependent variables (2 step GMM-levels estimation)
Dependent
variables
Loan growth
rate
NIM
NPL
ROA
Provisions/
Total Assets
Lagged
variable
0.39***
(8.4)
0.95***
(109.0)
0.813***
(53.4)
0.67***
(14.8)
0.473***
(4.6)
D(CPP)
0.074***
(3.1)
-0.0017*
(1.7)
-0.0076*
(1.9)
0.0033**
(2.0)
-0.0024*
(1.6)
No. Obs
4185
4180
2962
4182
4086
Joint Wald
330 [0.0]***
3900 [0.0]***
5577 [0.0]***
756 [0.0]***
1323 [0.0]
Sargan
427 [0.06]*
374 [0.62]
333 [1.0]
431 [0.4]
474 [0.5]
AR(1)
-3.07 [0.002]***
-0.72 [0.4]
0.028 [0.98]
-0.88 [0.4]
-1.4 [0.15]
AR(2)
1.04 [0.3]
-0.84 [0.4]
-0.83 [0.4]
-0.37 [0.71]
0.36 [0.72]
Conclusions
• Results indicate that commercial property
prices have a major impact on a wide range
of bank performance variables
• Signs found are consistent with a view that
commercial property provides important
forms of collateral perceived by banks to
reduce risk and encourage lending
• Results hold consistently across a number
of econometric specifications, as well as for
regions.
• Interesting differences in response of small and large
banks
– Commercial property price movements having a smaller
effect on the loan quality and provisions of small than large
banks
– Small bank profits less geared to commercial property prices
than are those of large banks. Consistent with large banks
being more willing to take risk as a consequence of the safety
net and moral hazard.
• Generally, results underline crucial relevance of
commercial property prices as macroprudential
variable. Need for good data on prices
• Also highlight the need to develop indicators of
individual bank exposure to the property market for
stress testing (note – wider than CP lending per se
given use as collateral)
References
• Davis E P and Zhu H (2004), "Bank lending
and commercial property prices, some cross
country evidence", BIS Working Paper No 150
• Davis E Philip and Haibin Zhu (2005),
"Commercial property prices and bank
performance", BIS Working Paper No 175 and
Quarterly Review of Economics and Finance,
49, 1341-1359